Signature File Methods for Text Data Bases (Computer and Information Science)

本研究对文本数据库的高效检索方法进行了设计和分析。它利用了一种名为签名文件的新方法，这种方法适用于更新非常活跃的档案应用程序。该方法特别适用于光盘。具体地说，通过分析和模拟将签名文件与其他索引方法进行了比较。实现并评估了几种可供选择的签名提取方法以验证分析结果，并开发了涵盖文本本身和关键字属性的集成签名方法。为了进行真实的实验，使用了一个原型检索系统，它存储和检索技术文章的摘要。马里兰大学的研究人员可以使用这个原型，以帮助他们进行文献搜索，并向系统设计者反馈需要哪些功能以及如何使用该系统。以发表的文章和报告的形式获取的科学知识数量巨大，而且还在不断增加。这个“文本数据库”太大了，如果没有计算机的帮助，就无法有效地使用。因此，研究自动检索方法，使计算机系统能够进行有效和广泛的检索，具有极其重要的科学和实际意义。

This research develops the design and analysis of efficient searching methods for text data bases. It utilizes a new method, called signature files, which works well for archival applications with very active updating. The method is especially suitable for optical disks. Specifically, signature files are compared with other indexing methods analytically and through simulation. Several alternative signature extraction methods are implemented and evaluated to verify analytical results, and integrated signature methods, which cover both the text itself and "keyword" attributes, are developed. To perform realistic experiments, a prototype retrieval system is used, which stores and retrieves abstracts of technical articles. This prototype is available to researchers at the University of Maryland, to assist them with literature searching and to give feedback to the system designers about what capabilities are needed and how the system is used. There is an enormous and increasing amount of scientific knowledge captured in the form of published articles and reports. This "text database" is far too large to be effectively used without computer assistance. Consequently, research on automatic retrieval methods, which allow effective and extensive searching to be done with computer systems, is of paramount scientific and practical importance.